laclong

LacLong is designed to sit between AI clients and various AI providers, acting as a unified gateway for managing provider contracts, routing decisions, and account selection. It aims to provide a single OpenAI-compatible endpoint to simplify the integration of multiple AI models and accounts. Currently in its foundation phase (Phase 0), the project provides the following capabilities: - A TypeScript-based gateway with a mock OpenAI-compatible `POST /v1/chat/completions` endpoint. - Basic health checks and model listing endpoints. - Implementation of priority and round-robin routing logic. - A Laravel-based dashboard foundation for managing provider, account, and usage data. - A credential management system using `CredentialVault` (supporting macOS Keychain or encrypted files). - Docker Compose integration for local environment setup. The architecture separates the gateway (Node.js/Fastify) from the administrative dashboard (Laravel), ensuring the gateway remains independent. Planned future features include production adapters for OpenAI, Anthropic, Gemini, OpenRouter, and Ollama, as well as streaming chat completions, quota tracking, and automatic failover.